00:02This was from the OpenAI agent hacking into a government portal system.
00:07It was a big deal at the time, the Prime Minister making the announcement in New York and revealing
00:11he'd had a conversation with the boss of OpenAI, expressing concern, although at the
00:17time and still saying that the impact of accessing these Medicare statistics was indeed minor.
00:24But the recommendations that's been delivered today, including strengthening engagement
00:29with Frontier Labs, encouraging AI labs to do more research here in Australia and training
00:37here in Australia.
00:38Another one is considering mandatory and time-bound notifications.
00:42So, you know, considering making these AI companies notify the government as quickly as possible
00:48about any bad behaviour.
00:49Considering bolstering cyber defences, although the government did stress today it felt like
00:55its cyber defences were strong.
00:57And then options to improve the criminal law as well, to ultimately improve AI safety, because
01:04that was a big question that emerged in the wake of this as well.
01:08Did the OpenAI agent, tongue twister there, break the law?
01:13And it appears from the comments that the Prime Minister made today that that wasn't the case.
01:17So, looking at making some changes there.
01:20But worth noting as well, we've got these 18 recommendations, but not actually the full
01:25report.
01:25The government says they'll release that one later on.
01:28All right.
01:28Is it known then what companies will actually be required to do under this proposed new regulator?
01:34Yeah.
01:34So, establishing a regulator, and that is going to have the power to impose big fines on companies
01:40that do the wrong thing.
01:42So, they'll be required to identify and manage serious risks associated with their models,
01:48and then allow for independent evaluation of their models as well.
01:53And then one of the other parts or powers that the regulator will have, that the government
01:58wants to eventually legislate, is also making them quickly notify the government of any incidents,
02:04because that was a big sort of sticking point in this breach with OpenAI as well.
02:10These protections will be regulated by a new targeted AI safety regulator, who will be able
02:18to enforce significant penalties for AI companies doing the wrong thing.
02:22Now, the Prime Minister said he wants to put these safety standards into law, and they will
02:27apply to all frontier companies that want to operate here in Australia.
